For American nations, the estimates of reporting factors were calculated empirically from the PAHO DENV cases dashboard PLISA Health Information Platform for the Americas. From June 2022 to February 2023 the PAHO DENV cases dashboard was downloaded weekly. Additionally the DENV case data for the same time period was downloaded from the PAHO DENV cases dashboard in July 2024. Using this data the reporting factor for each country at each lag in data reporting could be estimated using the following equation:

\[\mathbf{f}_{c,d} = (\frac{1}{T} \sum_{t=i}^{T} \left( \frac{N_{t,c,d}}{V_{t,c}} \right))/1\] where N is the number of DENV cases for a given country (c), at a given epiweek (t) for a given delay (d), V is the validated count of DENV cases for a given country (c), at a given epiweek (t) and f is the reporting factor for a given country (c), at a given epiweek (t) for a given delay (d). T is the number of observation recorded.

For all nations the most recent update was considered to be the most accurate data for the monthly counts of DENV cases. For PAHO nations this case count was multiplied by the average reporting factor at corresponding delay and country. Four nations (Belize, Dominica, Barbados and Paraguay) did not undergo this correct due to the complex nature of reporting factors in those countries. This corrections resulted in small differences in the monthly and overall cause count.

Understanding this underlying seasonal profile can provide a basis for predicting future case load, using observations of the number of cases observed to date within a given season. Dividing cumulative cases observed within a given season by the expected cumulative proportion returns an estimate of the total number of cases expected for that season. Multiplying this figure by expected monthly proportion returns an estimate of monthly cases. Fig. 2 uses data from Thailand and Fiji, locations with 32 and 3 seasons of data respectively, to demonstrate this use case. Here existing seasonal profiles are used to predict the most recent season with data available. Predicted and observed values are closer together for Thailand than Fiji, highlighting the increased predictive power provided by increasing the number of seasons used for prediction.

Using combined data from the OpenDengue and WHO dengue observatories we identify the calendar month with the lowest case load on average. We define this month as the first month of the dengue season. By example, if April is the calendar month with lowest case load on average, it becomes the first month of the season, and March the last month. This process normalises the seasonal profiles across different locations by setting a unified starting point. It also addresses instances where the peak season lies across the new year. Were analysis performed by calendar year, small changes in the timing of the season peak could lead to significant changes in the number of cases allocated to each year. Aligning data from calendar year to season accounts for such potential peak timing heterogeneity across years, which could skew results.

Once data has been aligned from calendar year to dengue season any locations with less than three seasons of data were removed. Monthly proportion of cases observed within each season was then calculated, normalising for between-season differences in total case load. Taking the average of this proportion across all seasons provides a baseline seasonal profile for each country. Standard deviation of this proportion provides a measure of uncertainty around this average proportion, accounting for between year differences in case load distribution. Figure 1 shows the average cumulative monthly proportion by country, with error bars representing the 95% confidence intervals.

CARD TEXT

based on cumulative data for the most recent month.

severity is defined by comparing the current season to the historical average season.